A NEW SPECIES OF MESOCLEMMYS, FROM THE OPEN FORMATIONS OF NORTHEASTERN BRAZIL (CHELONII, CHELIDAE)

A new species of South American side-necked turtle is described from the state of Piauí, Brazil. It is related to the Amazonian species of the genus Mesoclemmys s.l. (new acceptation) and is apparently a biogeographically relictual population of a species formerly living under a more humid climate in northeastern Brazil.


INTRODUCTION
Systematic knowledge of the Chelids (Chelidae), a turtle family widely distributed in the Australian and the Neotropical realms, is still progressing, with new species being brought to light sporadically.Among South American chelids, the genus Phrynops Wagler, 1830 (s.l.) is the least resolved taxonomically.A recent revision by McCord et al. (2001) recognizes six genera: Phr ynops Wagler, 1830, Rhinemys Wagler, 1830, Mesoclemmys Gray, 1863, Batrachemys Stejneger, 1909, Bufocephala McCord et al., 2001, and Ranacephala McCord et al., 2001.According to McCord et al. (2001), the genera Batrachemys and Phrynops include six and four species, respectively, whereas the genera Ranacephala, Rhinemys, Bufocephala, and Mesoclemmys are monotypical.Although we agree with the specific allocations of the genera Phrynops and Rhinemys proposed by these authors, we refrain to follow their taxonomic arrangements for the genera Ranacephala, Mesoclemmys, Bufocephala, and Batrachemys, which seem to be mainly based on two correlated characters, i.e., the width of the head and the degree of development of the 'parietal crest'.
Recent field works in northeastern Brasil revealed the existence of a previously unknown population of Chelid turtles here described as belonging to a new species attributed to the genus Mesoclemmys.The presence in this new species of the conspicuous external characters used by McCord et al. (2001) to establish the distinctiveness of Mesoclemmys, Batrachemys, Ranacephala, and Bufocephala reinforces the idea that such generic allocations are weakly supported by the morphological evidence at hand (see Appendix 3 and compare with Appendix B in McCord et al., 2001).Additionally, the lack of information on the internal anatomy of most species of 'toad-head turtles' contribute to perpetuate poorly supported taxonomic arrangements for the group.We tested the robustness of the phylogenetic scheme proposed by McCord et al. (2001) by including the new species described herein.Results of this test are shown below and support the view that a more cautious taxonomic arrangement than the one proposed by McCord et al. (2001) should be followed for the 'toad-headed' South American chelids.From a strictly nomenclatural point of view, the use of the oldest name Mesoclemmys to accommodate all species currently included in the genera Batrachemys, Bufocephala, Ranacephala, and Mesoclemmys sensu McCord et al. (2001) represents the most appropriate decision until a more robust phylogenetic hypothesis is proposed for the group (see Appendix 1 and Discussion below).

Distribution of the Genus Mesoclemmys
Geographical ranges of the ten species assigned here to the genus Mesoclemmys and discussed below are illustrated in Figure 1.Mesoclemmys nasuta (Guyanas), and above all M. dahli (Colombia) and M. zuliae (Venezuela), represent northern, isolated, and more or less relictual populations.Mesoclemmys raniceps is an Amazonian species which range extends from eastern Ecuador and Peru, northwestern Brasil, and northern Bolivia to Venezuela, and as far as the state of Pará in northern Brasil (Pritchard & Trebbau, 1984;Bour & Pauler, 1987;Iverson, 1992;McCord et al., 2001).Mesoclemmys heliostemma is limited to the northwestern part of this range (McCord et al., 2001).Mesoclemmys gibba has a wide distribution, rather similar to that of M. raniceps, with extensions on Trinidad and Guyanas (Mittermeier et al., 1978;Pritchard & Trebbau, 1984;Bour & Pauler, 1987;Iverson, 1992;McCord et al., 2001).Bour & Pauler (1987) noticed for M. gibba a geographical cline involving the color pattern, with individuals from the western part of the range (morph 'stenops' Spix, 1824) being duller than those of the eastern part (morph 'gibba').Recently, an isolated population provisionally refered to this species has been located in the central Brasilian state of Tocantins (Maran, 2004;Vetter, 2005;Fig. 1).Mesoclemmys hogei, M. tuberculata, and M. vanderhaegei are not Amazonian elements.Mesoclemmys hogei is limited to the Rio Paraíba drainage in Southeastern Brazil whereas M. tuberculata has a northeastern Brazilian range (corresponding with the Caatinga), and M. vanderhaegei has a Paraguayancentral Brazilian distribution (Bour & Pauler, 1987;Iverson, 1992;McCord et al., 2001).The latter two spe-cies are superficially similar and have sometimes been confused (see, e.g., McDiarmid & Foster, 1987), despite their distinct distribution ranges.

Serra das Confusões: Type-locality of the new species
Created in October 1998, the Parque Nacional da Serra das Confusões (PNSC) is located in the southeast of the state of Piauí (08°32'-09°16'S, 43°15'-43°51'W), covering an area of 5024 km², between the towns of Caracol, Guaribas, Cristino Castro, and Tamboril do Piauí (Fig. 2).The average altitude is close to 600 m; the climate is tropical semi-arid; the biome ecosystem is Caatinga.All the area drains into the Rio Parnaiba basin.Most of the area of the park corresponds to an extensive arenitic plateau, the "Chapada dos Gerais" (locally known as "Serra Grande"), dissected by an intermittent river drainage (Rodrigues et al., 2001;Zaher, 2002).The western and southern parts of the park are dominated by highly dissected rocky outcrops, forming a complex system of crevices and canyons that eventually open on lowland areas.The plateau is mostly covered by "Carrasco" vegetation (Fernandes, 2000) with an abundant leaf litter covering a sandy soil.The open lowlands, crevices, and canyons are covered by dry forest that grows on a sandy soil with small amount of leaf litter (Rodrigues et al., 2001;Zaher, 2002).
H. Zaher coordinated two field trips during the years of 2000 and 2002 to the PNSC in order to survey the terrestrial vertebrate fauna (Zaher, 2002).One of the interesting results was the survey of several species with obvious Amazonian affinities, notably a new species of Stenocercus Duméril & Bibron, 1837 (Iguania, Tropiduridae) and an isolated population of Vampyrum spectrum (Linnaeus, 1758) (Chiroptera, Phyllostomidae).This new species of Stenocercus was only found on the plateau of the "Chapada dos Gerais", with typical "Carrasco" vegetation and dry weather during most of the year.On the other hand, Vampyrum spectrum was collected inside the main canyon of the rocky area called "Olho d'Água da Santa" (Rodrigues et al., 2001) (Fig. 2).Here the climate is always wet and even in the hottest and driest season some waterholes remain; the temperature is rather stable.Outside these canyons, in the areas of open vegetation area, all the ponds and streams dry up during the hot season.Two species of chelid turtles were observed in association with aquatic habitats.Mesoclemmys tuberculata (Luederwaldt, 1926) was quite common and found nearly everywhere, includ-ing in the 'core area' of the park.That is because of its propensity to wander from one aquatic system to another, possibly at night (Vanzolini et al., 1980), and to bury itself during the dry time of year, a vital adaptation in this peculiar biome.It must be noted that this is apparently the first time that the species has been recorded from the Rio Parnaiba basin.
The second species of chelid turtle was found only on lowland areas near the arenitic outcrops, living near and inside the perennial waterholes present at the FIGURE 1. Schematic range of the recognized species of Mesoclemmys (from Bour & Pauler, 1987;Iverson, 1992;McCord et al., 2001).Interrogation mark represents an isolated population of Mesoclemmys cf.gibba.bottom of the canyons on the adjacent areas called "Baixão do Fausto" and "Olho d'Água da Santa (Fig. 2).Three specimens were collected.At first glance this 'canyon turtle' exhibited similarities with both Mesoclemmys vanderhaegei and M. gibba.A more detailed study, presently limited to the external characters, revealed that it was obviously a new species, by some features equally related to several other members of the genus Mesoclemmys (sensu lato).We propose to describe it as: Mesoclemmys perplexa sp.nov.
Figures 3 and 4, Table 1 Holotype: MZUSP 4111 (Museu de Zoologia da Universidade de São Paulo), a juvenile specimen (probably a female) with 67.6 mm of carapace length (see Table 1 for other measurements and Fig. 3), collected in 08 October 2000 by H. Zaher and team inside a small pond of the major canyon located in the region called "Olho d'Água da Santa", southern part of the Parque Nacional da Serra das Confusões.Etymology: Perplexa (latin) means confused, intricate, obscure, or ambiguous, and refers to the place where the taxon was observed, Serra das Confusões; it also alludes to its ambiguous generic attribution, and that of its allies.
Description of the holotype: Shell elongated; horizontal outline elliptical, barely notched at the level of the rear intermarginal sulci; longitudinal outline depressed.Dorsal keel smooth, continuous from V2 to V5. Areolae granular, wide; about five concentric growth rings, made of many small tubercles which draw radiating ridges.Supracaudal scutes subequal to M11.Plastron rather short and narrow; from the longest to the shortest median length of the scutes: intergular, abdominal, anal, femoral, humeral and pectoral.Intergular wider in front, its free border subequal to the gular border.Bridge short, its marginal sulcus curved, with distinct but small axillary (elongated) and inguinal (trapezoidal) scutes.Head flat, wide, and regularly pointed.Dorsal head scutes well delimited but not protruding, the frontal one making a cross-shaped figure similar to that of M. gibba.Skin of the throat slightly granular; two small and short barbels behind the lower rhamphotheca.Neck covered with numerous small, rounded projections, but without elongated or pointed tubercles.Limbs noticeably slender, slightly built.Front aspect of the forearm covered by about 4-5 longitudinal rows of oval, quadrangular or half-moon-shaped scutes, slightly overlapping.Legs with two rows of 4-5 enlarged scutes; the inside ones (fibular) half-moonshaped, the outside ones (tibial) rather rounded, and the most distal by far the larger.
Coloration: Dorsal aspect regularly dark, the shell brownish, the head, the neck and the limbs dull gray.Underside lighter; plastron yellowish, with a wide central brown symmetrical patch, extending from the humeral to the femoral, along the bridge, and pointing over the intergular, similar to the ornamentation shown by most  species of the genus Mesoclemmys here recognized (M.nasuta, M. gibba, M. heliostemma, M. raniceps, M. vanderhaegei, and juvenile M. tuberculata).Throat, tympanum and neck whitish, slightly and irregularly mottled with grayish spots; limbs and tail gray with limited whitish areas (arms, thighs, legs, tip of the tail).Laterally, on the head, the limit between the dark and the light areas is sinuous: starting from the nares, it passes through the maxillary horny sheath, then joins the corner of the mouth to the tympanum and follows its upper border.Apparently, the ocular dark stripe is absent.
Differences within the paratypes: MZUSP 4112, L = 100.8mm: its shell is very similar to the shell of the holotype; outline slightly more sinuous, lateral sides nearly parallel.Plastral median seams lengths are, from longest to shortest: intergular (straight length), femoral, abdominal, anal, pectoral, and humeral.The cruciform frontal figure is poorly delimited.The plastral color pattern is indistinct, covered by a superficial rusty stain.Whitish areas of the soft parts (limbs, neck) not so distinctly delimited; the throat widely covered by grayish spots.MZUSP 4086, L = 193.7 mm: outline distinctly constricted on the sides, partly because of a strong flaring of the posterior marginal scutes.Shell depressed, flat, but proportionally slightly higher than that of the juvenile; dorsal keel smoothed, still obvious on V3 and V4.Plastral median scute lengths are, from the longest to shortest: intergular, femoral, abdominal, humeral, anal, and pectoral; the short anal scutes are probably related to the sex of the specimen.The seam separating the plastron from the marginal scutes, along the bridge, is deeply grooved, as in M. gibba; axillary and inguinal scutes barely distinct.The underside of the marginals, the gular, and the femoral scutes are only slightly washed with a brown tinge, without any obvious dark flecks.Light areas are rather extended on the neck and the limbs; the throat is medially mottled by a patch of small grayish spots; a few similar spots cover the ventral side of the lower jaw.
Comparison with the other species of Mesoclemmys (Appendix 3): Mesoclemmys perplexa differs from all other species of Mesoclemmys by a narrower and a more depressed carapace, associated with a moderate parieto-squamosal arch and a narrow parietal roof (see Fig. 5).Addi-  (Luederwaldt, 1926:442-443).We also provide in Appendix 3 a discussion of the 19 characters proposed by McCord et al. (Appendix B, 2001:750), and their respective character states present in M. perplexa.
Quantitative analysis: We also performed a canonical discriminant analysis over seven logaritmized body measurements (maximum carapace width, maximum carapace depth, maximum plastron length, median plastron length, minimum bridge length, and maximum head width) for 18 adult females of the three exter-nally very similar species M. per plexa (n = 1), Mesoclemmys vanderhaegei (n = 3), and M. gibba (n = 14).Additionally, we performed the same analysis for the 21 juvenile and subadult specimens from both sexes (n = 2, n = 4, and n = 15, respectively).Distances among groups were significant for the adult females (Wilk's , F = 7.92, DF = 14/20, p < 0.0001) and also for juveniles (Wilk's , F = 5.87, DF = 14/26, p < 0.0001 for juveniles).Scores of the specimens on the first and second axes for both analyses are represented in Fig. 6.All statistical analyses were performed using SAS 9.0.
Biological data: Except for its unique biotope and its apparent dependence on perennial water bodies, nothing is known about the eco-ethology of this turtle.The apparent dependence to perennial waterholes, in contrast to the sympatric M. tuberculata, together with its isolated and restricted range, suggests that the species exhibits a relictual distribution from a previously wider geographical range established during a much wetter climatic episode of northeastern Brazil.

DISCUSSION
A parsimony analysis was performed using the TNT program (Goloboff et al., 2003) on McCord et al.'s (2001) original data matrix with the new species Mesoclemmys perplexa included in order to evaluate its phylogenetic position within the taxonomic scheme proposed by these authors (Fig. 7).We reviewed all codings for the 18 characters proposed by McCord  4, 5, 7, 9, 13, 17, and 18 for Chelus fimbriata, Mesoclemmys gibba, M. vanderhaegei, M. tuberculata, M. hogei, and Phrynops geoffroanus, based on our own observations (see data matrix in Appendix 4).We also added one new character to their data matrix [character 19 -Carapace width in adults with CW/CL superior to 72% (0), equal to 71-68% (1), inferior to 67% (2)].The character list and data matrix are offered in Appendix 4.
The analysis, implemented using the implicit enumeration (branch and bound) search option, resulted in 145 equally most parsimonious trees with a tree length of 49 steps (all multistate characters non-additive), an ensemble consistency index of 0.49 and a retention index of 0.58.The strict consensus tree supports a basal position of P. geoffroanus (Fig. 7) while Rhinemys rufipes appears as the sister-taxon to a clade formed by the remaining 'toad-headed' species, as previously suggested by McCord et al. (2001).On the other hand, the present analysis does not support the phylogenetic hypothesis suggested by McCord et al. (2001) for the remaining 'toad-head' turtles, the latter clade corresponding to a polytomy including their Mesoclemmys gibba, Bufocephala vanderhaegei, Ranacephala hogei, Batrachemys dahli, B. heliostemma, B. nasuta, B. raniceps, B. tuberculata, and B. zuliae, and the new species described in the present study (Fig. 7).Additionally, both clades are only weakly supported, with bootstrap percentages inferior to 50% and a Bremer support of one.These results suggest that the taxonomic scheme proposed by McCord et al. (2001) for the 'toad-headed' South American chelids, with the recognition of five new genera including two new ones, is not appropriate respect to the present knowledge of the group.In that sense, the more conservative approach used here (i.e., including these species in the older genus Mesoclemmys) represents a better solution until a well supported hypothesis of relationships is proposed for the 'toadheaded' South American chelids.

APPENDIX 1
New taxonomic arrangement adopted in the present study for the species formerly recognized as belonging to the genus Phrynops sensu lato.
An individually variable character.A tendency to a reduction of the neural bones (from 5 to 0, with disappearance of the first one, N1) is documented for all species, except for R. rufipes, for P. geoffroanus and its allies, and for Chelus.
10) Presence of first neural: Apparently absent.Cf. above, character (9).N1 is present in P. geoffroanus and its allies, also in Chelus fimbriata.

18) Upturned lateral borders of carapace:
A slight lateral constriction in the adult male.This character is significative to segregate P. geoffroanus, which has no upturned lateral borders.M. hogei is said to also have no such folded lateral marginals, but Luederwaldt (1926:443;pl.)described and figured a specimen of M. hogei with a "Carapaça, nos bordos lateraes, levantada para cima…" (his specimen number 96).

19) Plastral pattern:
A yellowish background with a large central dark (brown to blackish) area, covering most of the scutes between humerals and femorals, extending medially on the intergular, but not on the bridges; underside of marginals yellow.

FIGURE 2 .
FIGURE 2. Satellite view (NASA data) of Parque Nacional da Serra das Confusões.The arrows in insert A point to the localities known as "Olho d'Água da Santa" and "Baixão do Fausto", where the holotype and the paratypes were collected.
Type locality: Forested and humid areas of the regions called "Olho d'Água da Santa"and "Baixão do Fausto", southern part of the Parque Nacional da Serra das Confusões, State of Piauí, Brasil.Paratypes: MZUSP 4112, a subadult female with 100.8 mm of carapace length (Fig. 4A, B), collected the same day and at the same location as the holotype.MZUSP 4086, an adult female with 193.7 mm of carapace length (Fig. 4C, D), collected in 14 January 2002 by H. Zaher and team when crossing an open area at night in the region called "Baixão do Fausto", southern part of the Parque Nacional da Serra das Confusões.

FIGURE 6 .
FIGURE 6. Scores of the adult females (A) and juveniles (B) specimens of Mesoclemmys gibba, M. vanderhaegei, and M. perplexa over the first and second axes of the canonical discriminant analysis.

TABLE 1 .
Holotype and paratypes of Mesoclemmys perplexa: main measurements, with proportions (in relation to the length of the shell)., M. perplexa differs from all other species of Mesoclemmys, except M. gibba, by the presence of a moderate keel in the carapace (a flat or even a medially depressed shell when they age).It differs from M. dahli, M. raniceps, M. vanderhaegei, and M. zuliae by the lack of a dark stripe crossing the eye (obvious dark head stripes are present in M. dahli, M. raniceps, M. zuliae; a more or less dark stripe crossing the eye can be observed in M. dahli, M. vanderhaegei, and M. zuliae).Mesoclemmys perplexa differs from M. dahli, M. hogei and M. zuliae by a pigmented plastron.Mesoclemmys perplexa differs from M. hogei and M. vanderhaegei by the lack of prominent temporal scutes.Finally, it differs from M. hogei by an obviously wider head, thinner barbels, a shorter median abdominal seam.Besides pointing out the identity of M. perplexa, this study led to transfer M. hogei and M. vanderhaegei from their monotypic genus Ranacephala and Bufocephala, respectively, to the genus Mesoclemmys in its present acceptation.We can notice that Luederwaldt had already proposed this generic attribution to specimens presently identified as M. hogei some 80 years ago FIGURE 5. Morphometric differences between Mesoclemmys gibba, M. vanderhaegei, and M. perplexa: ratio Carapace length/Carapace depth (CL/CD) versus ratio Carapace width/Head width (CW/ HW).tionally Type species: by monotypy: Emys geoffroana Schweigger, 1812.Type species: by subsequent designation: Emys rufipes Spix, 1824.Bufocephala McCord et al., 2001 (type species, by original designation: Phrynops vanderhaegei Bour, 1973); Ranacephala McCord et al., 2001 (type species, by original designation: Phrynops hogei Mertens, 1970).
gibba, M. vanderhaegei, and M. tuberculata, and the tectiform one of R. rufipes.Lowest shells are those of M. perplexa (28%, one adult female), and of some M. hogei, M. nasuta, M. raniceps and probably M. heliostemma (being slightly under 29%).14)Presence of carapacial keel:Yes, moderate, smooth.A slightly stronger keel, broken off, is typical of M. gibba.Other adult members of the group have a flat of even a medially depressed shell when they age.The tectiform shell of R. rufipes is quite peculiar.15)Presence of median groove in carapace: